MCP Code Executor

Example Use Cases

• 1

  Run arbitrary Python snippets in a reproducible environment (Conda/venv) and capture outputs.

• 2

  Build large or multi-part Python programs incrementally (initialize, append, then execute) to avoid LLM token limits.

• 3

  Install or check Python dependencies at runtime and execute scripts that rely on those packages.

Description

The MCP Code Executor runs Python code produced by LLMs inside a specified Python environment (Conda, venv, or UV venv), allowing execution with access to installed libraries and dependencies. It supports installing and checking packages, configuring the runtime environment dynamically, and storing generated code in a configurable directory. For large code blocks it provides an incremental workflow (initialize, append, verify, execute) to avoid token limits. This enables agents to produce, manage, and run multi-file or long scripts reliably within a controlled environment.